Originally published as an Advisory Hour Substack post on 2026-09-10. This first-party site copy preserves the text, localizes public supporting media, and connects the record to its project trail.
Original post
Visitors to Gamelab may have noticed the new change. The reason for the change is growth. Gamelab grew awkwardly over the year. It’s the departure zone for many experiments. It houses many of the AI related competition entries. For a time, it was fine. Then the site outgrew itself and a pondering of a new article on information architecture had me realize that it was time to come in and lay out a revamp.
The AI imaging tools that accept image inputs are underrated for idea exploration. The above vertical tower of the build was something I tossed into ChatGPT’s app and merely asked, “Let’s imagine new ways to organize this information”. I was in search of an idea benefiting the concept of gamelab, and also yet somehow representative of the overall thesis of my site: discovery thru action. A few designs back and forth and I just really enjoyed the taxonomy split of this concept. It made the gamelab feel more like a gamelab and less like an informational yellowpages.
The trick was getting the model to start thinking in terms of mission and taxonomy. Once those two bigger concepts started a delicate dance, then I found the result quality of what I was seeing off of the model improved dramatically. It wasn’t now a function of “yet another large listing page”. Instead, it was a container of more relevant content presented in a way that mechanistically just fit with the broader ethos of the site: building things is a fun way to FAFO.
Astra is such a great model for these context-heavy workloads. My site is extreme in this regard. There’s a volumes of content covering so many different projects and experiments. I like making things. The invisible hand at work in my projects isn’t always visible to me. However, hidden signals are visible to a model as capable as OpenAI’s Astra. So much so that the model was able to construct a fun lineage of projects to epxlore!
While the listing here is not a chronological lineage, the projects follow a concept lineage. This is arguably way cooler. Each project contains a question. This is my suggestion of a practice of Philosophy for the everyday man like me. Questions are asked and then understood by way of worlds. The answer isn’t a binary determination or subtle wordplay.
The atlas view is a representation of the projects in a composite view. Concepts and questions are linked and organized into a refined taxonomy from the old “list it all and figure it out later” approach. I let the body of work, the ever growing list of whatever project interests me, become the compass itself.
I share this all with you with a word of warning and sense of urgency: go create now. I encourage you to stack projects and create to such an extent that no immediate sense of organization or hidden hand is visible inside the day-to-day work that you undertake. Start now because the future needs you. In time, you’ll discover a treasure far richer than a pile of gold. There’s creative consequence at this scale.
Don’t miss out by putting it off.
Connected work
- GameLab Project Lineage: Follow the conceptual links among the experiments in the live lineage view.
- Information Architecture: Using Taxonomy With AI-Driven Results: Read the earlier note about the word that helped an agent organize a crowded menu.
- Frontier Atlas: Search the wider site by topic, project, format, and status.